Recent application domains such as the Internet of Things (IoT) and Smart Cities (SCs) have introduced novel challenges to Cloud Computing based on their stringent requirements (e.g., low latency, high bandwidth). With the exponential growth of IoT traffic in the last few years, traditional cloud systems have become inadequate for these applications since requests are made on-demand simultaneously by multiple devices at different locations. The Fog Computing (FC) paradigm has emerged to deal with the limitations of traditional clouds since computational resources are placed at the edges of the network, aiming to decrease the latency expected by IoT devices and reduce the amount of data sent to the cloud. However, research challenges persist in FC since it is not a mature concept yet. This PhD research addresses four challenges in the FC domain focused on providing an efficient resource allocation in these distributed infrastructures. This dissertation includes theoretical formulations as benchmarks for resource allocation, fog-based architectural concepts, anomaly detection practices for IoT, and latency-aware allocation approaches that lead to the implementation of a network-aware framework named Diktyo. It optimizes the allocation of container-based service chains by considering latency and bandwidth in the scheduling process of a well-known container orchestration platform, Kubernetes. Ex-periments showed thatDiktyo increases throughput by 22% and reduces latency by 45% for microservice benchmark applications.